🚀 go-pugleaf

In gdb 3.5, for decstations, the macro for INVALID_FLOAT defined in
m-mips.h is not correct.  The macro is as follows:

#define INVALID_FLOAT(p, len) ((*(short *) p & 0xff80) == 0x8000)

For floats on a mips, the floats are 1 sign bit, 8 exponent bits, 23
mantissa bit (with the sign bit most significant).  But, the above
macro, since decstations are byte-wise braindead (or vax-like), the
above bit operation is happening somewhere in the mantissa.  The
following is what happens:
	(gdb) print (float)331
	Invalid float value

To fix this, I think (though I wouldn't bet any amount) that the
following macro is a correct fix:

#define INVALID_FLOAT(p,len) ((*(int *)p & 0xff800000) == 0x80000000)

Worst case, you could just make the macro into the degenerate version
(which many systypes seem to have):
#define INVALID_FLOAT(p,len) 0

Oh yeah, you probably don't want to just change the m-mips.h file, but
make a separate file for decstations and Mips machines.

- Doug DeCarlo